This new version should have been the only version present in graphics cards. Nvidia was in a hurry to put it to use. When it was already a part of cards in client PCs, it became kind of useless to study its properties anymore. Until problems appeared, that is. Let's hope the industry learned something from this.

Neither did the 12vHpwr in lab testing. [youtube=7kyS0E0B2DU]

So in the end it was REALLY the connector... lol

Nope. Was end users and snakeoil adapter manufacturers.

Kaarme:

This new version should have been the only version present in graphics cards. Nvidia was in a hurry to put it to use. When it was already a part of cards in client PCs, it became kind of useless to study its properties anymore. Until problems appeared, that is. Let's hope the industry learned something from this.

Since this new version came about because of the issues detected with the 12VHPWR connector, it would have been hard for NVIDIA to adopt this connector initially. Also the 12VHPWR connector was pushed by the PCI-Express SIG, which determines all things related to the PCI-Express standards. The original design for the 12VHPWR connector was passed to the PCI-E SIG by Intel BTW, so if someone is to blame it is Intel for creating such a shoddy design in the first place. Was NVIDIA a bit too eager to move to the new adapter? Maybe, but I can see why it appealed to them: more power available to them in a smaller form factor. Remember in order to push 600 Watt (what they originally expected would be necessary for the RTX 4090 before they found out that the power consumption for Ada Lovelace was actually much lower than expected) they would have had to add four 8 pin PCI-E old style connectors, which would have required a tremendous amount of PCB real estate! TL;LR: So Intel designed a shoddy connector, which the PCI-E SIG did not examine thoroughly enough during their evaluation phase before adopting it as a new standard, which NVIDIA was maybe too eager to adopt when it became available because it solved one of the issues they were afraid they needed to deal with.

Crazy Joe:

it would have been hard for NVIDIA to adopt this connector initially. Also the 12VHPWR connector was pushed by the PCI-Express SIG

Wasn't hard at all, rolling production change and the new socket was in use on first runs of the 4070/ti

Crazy Joe:

TL;LR: So Intel designed a shoddy connector, which the PCI-E SIG did not examine thoroughly enough during their evaluation phase before adopting it as a new standard, which NVIDIA was maybe too eager to adopt when it became available because it solved one of the issues they were afraid they needed to deal with.

Intels testing of the connectors were solid, what they did not account for was third party deviation away from a female terminal that had contact on all 4 sides the entire length of the pin, being substituted with one that had the contact by 2-3 dimples on 2 sides of it.

Intel's own graphics cards use the old connectors. So, Intel itself didn't trust the new one. PCI-E SIG is basically nothing concrete, just a standardisation organisation. It doesn't have any physical products on store shelves, so it neither cares to nor can validate anything physical from real life industrial production point of view. It can only stare at numbers on a computer screen and say something should work, which sometimes doesn't turn out to be true. That was exactly the pitfall in this case.

Astyanax:

Wasn't hard at all, rolling production change and the new socket was in use on first runs of the 4070/ti Intels testing of the connectors were solid, what they did not account for was third party deviation away from a female terminal that had contact on all 4 sides the entire length of the pin, being substituted with one that had the contact by 2-3 dimples on 2 sides of it.

3-dimple design seems to be sufficient according to many PSU manufacturers. And 4-spring currently doesn't offer better contact with the wattage that's being pulled now, that's what they say atleast.

Jensen to team: Our beta testing of the 12VHPWR is now concluded. Please proceed with our final and improved release of 12V-2x6.

Astyanax:

Nope. Was end users and snakeoil adapter manufacturers.

Should I be ok using the official seasonic pcie5 cable for the titanium range ?

Astyanax:

Wasn't hard at all, rolling production change and the new socket was in use on first runs of the 4070/ti

I guess you missed the word "Initially" there. The OP said that NVIDIA should have used the new connector from the beginning i.s.o the 12VHPWR connector.

Astyanax:

Intels testing of the connectors were solid, what they did not account for was third party deviation away from a female terminal that had contact on all 4 sides the entire length of the pin, being substituted with one that had the contact by 2-3 dimples on 2 sides of it.

No, the issue was that the power pins could be partially or not connected while the sense pins were connected. This is what caused the overheating issues as there was arcing happening between the GPU side of the power socket and the cable side because the sense pins were telling the PSU to send the power even though there was limited or no connection of the power pins. The new connector solves this by shortening the sense pins so that the PSU only detects that it can send a particular power level through the power pins when they are fully connected. Since Intel didn't test the scenario of sense pins connecting with the power pins not or partially connected, they didn't fully test their design. You always have to check for all possible scenarios when it comes to these kinds of new connectors.

WAIT... i'm now confused, there was the origional 1VHPWR, then I heard about an new one which was going to use two x 6 connectors basically a 12VHPWR with some tweaks cut in half two make two connectors and now there is another tweaked single connector?? What I want to know which no one seems to be saying..Soes this mean there is a new cable that will stop this happening for existing users or is this only from now with new cards being produced... either way it feels like those who bought there cards befire this new connector are being screwed over.

VaultDweller:

WAIT... i'm now confused, there was the origional 1VHPWR, then I heard about an new one which was going to use two x 6 connectors basically a 12VHPWR with some tweaks cut in half two make two connectors and now there is another tweaked single connector?? What I want to know which no one seems to be saying..Soes this mean there is a new cable that will stop this happening for existing users or is this only from now with new cards being produced... either way it feels like those who bought there cards befire this new connector are being screwed over.

12V-2x6 is the same as 12VHPWR, just some pins lengths are adjusted, it's compatible with all current 12VHPWR cables on the market.

Astyanax:

Nope. Was end users and snakeoil adapter manufacturers.

Do you live in an alternate reality or is my memory screwed? All the videos from reputable sources I saw, like from GN, showed clearly that the connector is the problem.

Neo Cyrus:

Do you live in an alternate reality or is my memory screwed? All the videos from reputable sources I saw, like from GN, showed clearly that the connector is the problem.

@Astyanax is far smarter than those guys :P

TheDeeGee:

3-dimple design seems to be sufficient according to many PSU manufacturers. And 4-spring currently doesn't offer better contact with the wattage that's being pulled now, that's what they say atleast.

Sure, when they are plugged in all the way and the terminal hasn't split due to an angled insertion. The spring terminals just have more tolerance for doing stuff wrong.

Crazy Joe:

No, the issue was that the power pins could be partially or not connected while the sense pins were connected. This is what caused the overheating issues as there was arcing happening between the GPU side of the power socket and the cable side because the sense pins were telling the PSU to send the power even though there was limited or no connection of the power pins.

No, this wasn't the "issue", that the testing above in the image is looking into The testing in the image above is only being done at conditions where the sense pins are still in contact, they aren't testing for the disengage factor. The terminal heatup factors were always about poor contact, but not always in a way that even the 12v-2x6 would stop the power flowing. matching the conditions between 12hpwr and 12v-2x6 in the above tests, the 12hpwr wouldn't have overheated either, except in conditions of defect or the case of CableMods brainwave 2 dimple design that is inferior to the reference specs alternative.

Undying:

@Astyanax is far smarter than those guys 😛

I thought everybody knew that? I heard he even knows how a flux capacitor works? 😛

Astyanax:

No, this wasn't the "issue", that the testing above in the image is looking into The testing in the image above is only being done at conditions where the sense pins are still in contact, they aren't testing for the disengage factor. The terminal heatup factors were always about poor contact, but not always in a way that even the 12v-2x6 would stop the power flowing. matching the conditions between 12hpwr and 12v-2x6 in the above tests, the 12hpwr wouldn't have overheated either, except in conditions of defect or the case of CableMods brainwave 2 dimple design that is inferior to the reference specs alternative.

I wasn't talking about the new connector and the testing shown in the image, I was talking about why the original 12VHPWR connector failed, which you just reiterated. Seems hard for you to read what others are writing before you comment on it, doesn't it.

I just got a 4070 and I'm glad it's only got the good ol 8 pin. And only 1 to boot.

Astyanax:

Sure, when they are plugged in all the way and the terminal hasn't split due to an angled insertion. The spring terminals just have more tolerance for doing stuff wrong. No, this wasn't the "issue", that the testing above in the image is looking into The testing in the image above is only being done at conditions where the sense pins are still in contact, they aren't testing for the disengage factor. The terminal heatup factors were always about poor contact, but not always in a way that even the 12v-2x6 would stop the power flowing. matching the conditions between 12hpwr and 12v-2x6 in the above tests, the 12hpwr wouldn't have overheated either, except in conditions of defect or the case of CableMods brainwave 2 dimple design that is inferior to the reference specs alternative.

I have a feeling the reason those CableMod adapters and even cable extensions melt is because 12VHPWR cannot handle more than 2 connections. It would actually benefit from not being a modular cable.